# Open Ethernet Drive: Evolution of Energy-Efficient Storage Technology **Authors:** Hariharan Devarajan, Anthony Kougkas, Hsing Bung Chen, Xian-He Sun **Venue:** The ACM SIGHPC 8th International Workshop on Data-Intensive Computing in the Clouds (DataCloud 2017), in conjunction with SC’17, Denver, CO, USA, vol. 17 (2017) ## Abstract An Open Ethernet Drive, also known as an OED, is a new technology that embeds a low-powered processor, a fixed-size memory and an Ethernet card on a disk drive (e.g. HDD or SSD). All major storage vendors have introduced their respective implementations with similar architectural designs. As the technology progresses into its second generation, the performance characteristics have improved substantially. In this study, we first demonstrate the differences between two generations of the OED technology. We run a variety of benchmarks and applications to thoroughly evaluate the performance of this device and its compatibility with the current ecosystem. Furthermore, we investigate the performance and energy footprint of the OED technology when used as a storage server and as an I/O accelerator. Evaluation results show that OED technology can be a reliable, scalable, energy and cost efficient storage solution. It is a viable replacement for storage servers offering competitive performance while consuming only 10% of the power that a typical storage node needs. Finally, this study shows that OED’s can be used as I/O accelerators capable of executing data-intensive computations (such as sorting, compression/decompression, etc.) on local data, whereby the expensive data movement is minimized resulting in low power consumption. ## Links - [Paper](https://akougkas.io/assets/pdf/devarajan2017openethernet.pdf) ## BibTeX ```bibtex @inproceedings{devarajan2017openethernet, author = {Devarajan, Hariharan and Kougkas, Anthony and Chen, Hsing Bung and Sun, Xian-He}, title = {{Open Ethernet Drive: Evolution of Energy-Efficient Storage Technology}}, booktitle = {The ACM SIGHPC 8th International Workshop on Data-Intensive Computing in the Clouds (DataCloud 2017), in conjunction with SC'17, Denver, CO, USA, vol. 17}, year = {2017}, month = nov, publisher = {ACM}, url = {http://cs.iit.edu/%7Escs/assets/files/OED_DataCloud17.pdf}, } ```
# Open Ethernet Drive: Evolution of Energy-Efficient Storage Technology - **Source:** https://akougkas.io/publications/devarajan2017openethernet/ - **Authors:** Hariharan Devarajan, Anthony Kougkas, Hsing Bung Chen, Xian-He Sun - **Venue:** The ACM SIGHPC 8th International Workshop on Data-Intensive Computing in the Clouds (DataCloud 2017), in conjunction with SC’17, Denver, CO, USA, vol. 17 (2017) - **Summary:** An Open Ethernet Drive, also known as an OED, is a new technology that embeds a low-powered processor, a fixed-size memory and an Ethernet card on a disk drive (e.g. --- ## Abstract An Open Ethernet Drive, also known as an OED, is a new technology that embeds a low-powered processor, a fixed-size memory and an Ethernet card on a disk drive (e.g. HDD or SSD). All major storage vendors have introduced their respective implementations with similar architectural designs. As the technology progresses into its second generation, the performance characteristics have improved substantially. In this study, we first demonstrate the differences between two generations of the OED technology. We run a variety of benchmarks and applications to thoroughly evaluate the performance of this device and its compatibility with the current ecosystem. Furthermore, we investigate the performance and energy footprint of the OED technology when used as a storage server and as an I/O accelerator. Evaluation results show that OED technology can be a reliable, scalable, energy and cost efficient storage solution. It is a viable replacement for storage servers offering competitive performance while consuming only 10% of the power that a typical storage node needs. Finally, this study shows that OED’s can be used as I/O accelerators capable of executing data-intensive computations (such as sorting, compression/decompression, etc.) on local data, whereby the expensive data movement is minimized resulting in low power consumption. ## Canonical links - [Paper](https://akougkas.io/assets/pdf/devarajan2017openethernet.pdf) ## Citation ```bibtex @inproceedings{devarajan2017openethernet, author = {Devarajan, Hariharan and Kougkas, Anthony and Chen, Hsing Bung and Sun, Xian-He}, title = {{Open Ethernet Drive: Evolution of Energy-Efficient Storage Technology}}, booktitle = {The ACM SIGHPC 8th International Workshop on Data-Intensive Computing in the Clouds (DataCloud 2017), in conjunction with SC'17, Denver, CO, USA, vol. 17}, year = {2017}, month = nov, publisher = {ACM}, url = {http://cs.iit.edu/%7Escs/assets/files/OED_DataCloud17.pdf}, } ```